Global Optimization Methods And Superstructure Model For Heat Integration Of Heat Exchanger Networks

Heat integration of heat exchanger network(HEN)is the important part in process system synthesis,which can effectively control the energy cost and improve the energy efficiency in the system.From the perspective of optimization,the simultaneous synthesis of HEN belongs to a mixed integer and nonlinear programming problem.The nonlinearity of system would be heavened with the increase of stream quantity,increasing the difficulty of global optimization for HEN.The recent global optimization methods could not reconcile the optimization efficiency and the accuracy of the optimal solution while the most widely used stage-wise superstructure(SWS)model might play a restricted role in optimization methods and bring about the obstacle in realization of global optimization for complex HEN problems.Therefore,from the perspectives of global optimization methods and superstructure model,this paper studied and developed novel global optimization algorithms with high efficiency and accuracy,as well as novel superstructure model for heat exchanger network,to overall enhance the solution accuracy and optimization efficiency in heat integration of HEN.The major contents and contributions in this paper are stated as follows:Firstly,the optimization mechanisms and characteristics of stochastic methods were constracted and concluded and then the nature in the prematural convergence was revealed taking particle swarm optimization(PSO)as an example.Moreover,a forced jump-out strategy was proposed to update the optimal search direction by random variables around the local optimum.The results demonstrated that the improved PSO could effectively enhance the accuracy in continuous variables optimization and search for better HEN structures by continuously jumping out of local optimum.Secondly,a novel random walk algorithm with compulsive evolution(RWCE)was first established to realize the simultaneous optimization for integer and continuous variables in SWS.Moreover,the mechanism of the productive global search ability in RWCE was revealed from the perspectives of algorithm mechanism and clustering phenomenon.Further,fine-search strategy and ramdom perturbation strategies were established,thus an integrated RWCE algorithm system with local optimization accuracy and global optimization efficiency was formed,applicable to heat integration of large-scaled HEN problems.(I)RWCE: RWCE was established for the first time on the basis of pure random evolution mechanism,compulsive evolution principle and mechanism of accepting the imperfect solution with small probability,and successfully applied to the global optimization for HEN without stream splits,realizing the simualtaneous optimization for integer and continuous variables.Further,the effects of key parameters in RWCE on the performance of algorithm were analyzed.The research demonstrated that RWCE possessed the merits of productive operability,adaptability and the ability of keeping the individuals' vitality during the evolution,thus had great global search ability.(II)FS-RWCE: Fine-search strategy was first established to improve the nonlinear optimization performance and enhance accuracy of local search in RWCE.Then an integrated random walk algorithm with compulsive evolution and fine-search strategy(FS-RWCE)was established and successfully applied to the global optimization for large-scaled HEN problems.The results indicated that FS-RWCE could effectively enhance the accuracy of continuous and integer varables optimization,thus to reconcile the local optimization accuracy and global optimization efficiency.(III)FS-RWCE with random perturbations: Random perturbation strategy was first established to improve the performance of integer variables optimization and further enchance the efficiency of integer variables optimization in FS-RWCE.The effectiveness of random perturbation strategy was demonstrated by some case studies with different scales.The results showed that the optimization performance of continuous and integer variables for large-scaled HEN problems could be both enhanced by introducing the random perturbation strategy.(IV)Cluster effect in RWCE: The clustering phenomenon in the evolution process of RWCE was first revealed.Firstly,main evaluation indexes for structural similarity were investigated to identify and monitor different clusters with corresponding structural similarity automatically.Then,the characteristics of clustering phenomenon were concluded by observing the clusters distribution during the evolution.Moreover,the positive and negative effects of clustering phenomenon on the RWCE optimization performance were concluded by observing changes of the located clusters and the corresponding structural similarity for the optimal individuals and free individuals during the evolution.Finally,some attempts making full use of cluster effect were carried on to further enhance the global optimization performance.The research demonstrated that the combined role of structural similarities and free individuals facilitated the productive global search ability in RWCE.Enhancing the cluster effect by taking the optimal individual as the structural similarity of population and weakening the cluster effect by random perturbation could effectively reconcile the global and local search ability.Thirdly,a novel node-wise non-structural model(NW-NSM)was first established where the heat exchanger location was quantized by nodes at each stream and a HEN structure was formed by random match between hot and cold nodes.Applying RWCE to optimize NW-NSM,higher accuracy and efficiency could be achieved in the heat integration for large-scaled HEN problems.Moreover,uniform distribution strategy for heat exchangers was established to increase the probability of generating new stream matches.The results showed that the presented NW-NSM possessed more flexibility and freedom to expand the search region for feasible solutions.The combination of RWCE and NW-NSW could fully enhance the solution accuracy and optimization efficiency.Introducing uniform distribution strategy for heat exchangers could further enhance the optimization effiency for integer variables and expand the lead of NW-NSM in heat integration for HEN problems.In brief,this paper creatively proposed RWCE and NW-NSM from the two lines,that is,global optimization method and simultaneous synthesis model,and realized the effective optimization for integer and nonlinear variables.The establishments and combination of RWCE and NW-NSM became the breakthrough for solving the dilemma of global optimization for large-scaled HEN problems and exploited new directions for global heat integration for large-scaled and complex systems.